Test that complext and fast text codepaths measure the same width
Created attachment 241310 [details] Patch
I suspect that this test would have passed even with a change such as the one introduced by <http://trac.webkit.org/r175813>, because tests run in an environment where kerning and ligatures are disabled by default. Another good way to test that the two code paths agree is to use the --complex-text option of run-webkit-tests.
You're right; the test passes. This whole "not testing what we ship" thing is pretty terrible.
Should we add a way that a test can opt into the complex text path?
This isn't actually about the fast vs slow code paths, it's about how kerning and ligatures are turned off. Instead of opting in, I think it's time we enabled K&L for all tests by default.
(In reply to comment #5) > This isn't actually about the fast vs slow code paths, it's about how > kerning and ligatures are turned off. Instead of opting in, I think it's > time we enabled K&L for all tests by default. I agree. (Note that we do have some tests that enable kerning and/or ligatures using the relevant CSS properties, and we even have some tests that force the complex code path to be taken).
Comment on attachment 241310 [details] Patch Clearing flags on attachment: 241310 Committed r175843: <http://trac.webkit.org/changeset/175843>
All reviewed patches have been landed. Closing bug.